Will software Suck ?

· ai, aislop, future of software

Lately my feed has been overflowing with layoffs and news about companies have stopped hiring juniors. The mid & senior folks they kept get handed the juniors' work on top of their own. Why? Cuz AI makes everything fast.

My take is that we need the same rate of hiring growth we had before AI, if not more. Good software gets built by handing small pieces of work to people who'll craft them carefully, an optimisation here, a small feature there, lil research about a slightly faster or efficient algorithm or data struct, and the quality of the whole is the sum of that care. It's the classic quality vs quantity trope, and software tech is doubling on the everything opposite of it. ( this is not a generalization , rather what the trend is )

Look at this tomfoolery called tokenmaxxing.

Back in April 2026 it came out that Meta had an internal leaderboard, Claudeonomics, ranking 85,000+ employees by how many AI tokens they burned. The top burners got titles like Token Legend and Session Immortal. Can't make this stuff up. Meta, OpenAI & Microsoft all had some version of it, and some engineers reportedly set bots running in loops to climb the board while producing nothing. Of course they did. Measure people by how much they consume and they'll consume.

Meta took the board down eventually, but the pressure to show the AI bill is paying off is still very much there, and the software being churned out under it is a ticking bomb of CVEs. Folks at Georgia Tech tracking vulnerabilities traced back to AI coding tools counted 35 in March 2026 alone, up from a couple a few months before, and they reckon the real number is 5–10x that.

The bigger problem is the pipeline.

If the industry produces no juniors today, it'll have no seniors tomorrow.

You can't skip a whole tier of the pipeline and expect competence to show up out of thin air. A senior engineer is years of learning, breaking prod (hopefully not too often), getting corrected and soaking up context. Last I checked, there's no npm install senior-engineer. Take away the years and you take away the engineers.

Meanwhile the seniors, on top of the juniors' work, are reviewing code, designing systems and cleaning up after AI suggestions that sound extremely sure of themselves and fall apart the moment you look at the structure. And the industry's response? Move faster. Oh, and the models writing that stuff are increasingly trained on older AI output, so it's AI learning from AI, and that loop is already running.

A model's context window has practical limits. An engineer who's spent months inside a large codebase carries its history in their head, how it breaks, why it's built the way it is, which decisions were deliberate & which were 2am hacks that stuck around. You can't rebuild that from a context window, at least not with the LLMs we have today. When that engineer leaves, all of it leaves with them.

So if the industry keeps overloading the experienced folks it has left and refuses to grow new ones, it's not hard to see where this goes. Slower systems, compounding tech debt, and a generation of devs stuck maintaining & debugging codebases nobody around them fully understands, cuz the people who did are gone and never handed any of it over.

The money saved today comes with an invoice tomorrow. And it'll arrive with interest.

Older: Build || !Build
Newer: The Skyscraper Inside Your Program

A mug of coffee, still warm.